How Citizens and Their Legislators Prioritize Spheres of Representation
*
John D. Griffin† Associate Professor Department of Political Science University of Notre Dame
[email protected]
Patrick Flavin Assistant Professor Department of Political Science Baylor University
[email protected]
Forthcoming, Political Research Quarterly
*
A previous version of this paper was presented at the annual meeting of the Midwest Political Science Association, April 12-15, 2007, Palmer House Hilton, Chicago, IL. Barry Burden, Brandice CanesWrone, Jamie Druckman, Martin Gilens, Tobin Grant, David Kimball, Brian Newman, Benjamin Page, and Chris Wlezien provided helpful comments. We thank Adam Joines for research assistance. †
Corresponding author. Department of Political Science, 217 O’Shaughnessy Hall, University of Notre Dame, Notre Dame, Indiana 46556.
Abstract The authors uncover evidence that citizens’ priorities about various spheres of legislative representation differ across demographic groups and that these differences are subsequently reflected in the in-office behavior of their elected officials. Specifically, African Americans and Latinos are less concerned than whites with policy representation – the attentiveness of elected officials to citizens’ policy preferences – but place more emphasis on their district receiving its fair share of federal money. Citizens with higher incomes place a higher priority on policy representation and less on constituency service than do those with lower incomes. Importantly, these priorities map onto their member of Congress’s behavior.
A fundamental concern for American democracy is that citizens’ political preferences are weighed equally by their elected officials (Dahl 1971). However, a growing body of empirical work has demonstrated that traditionally disadvantaged groups such as the poor and racial minorities are less likely to have their political opinions represented by policymakers compared to the wealthy and whites (Bartels 2008; Gilens 2005; Griffin and Newman 2008). These studies of policy representation implicitly assume that demographic groups do not differ in how they prioritize various spheres of representation – that groups equally prioritize the representation of their political opinions by policymakers, for example. However, there are reasons to believe that some groups give greater weight than other groups to other spheres of representation such as securing federal money for one’s district or state or providing constituency service. If citizens prioritize spheres of representation differently, and elected officials are attentive to these differences, this casts representation inequalities in a new normative light. In this paper, we examine the nexus of citizens’ priorities concerning various spheres of representation and the in-office behavior of their Members of Congress (MCs). First, we show that African Americans and Latinos place a lower priority than whites on the fit between their own preferences and their MCs’ roll call decisions. Instead, racial and ethnic minorities place more emphasis than whites on their district receiving its fair share of federal money. We also find that citizens at higher income levels place a greater priority on policy representation and are less concerned about constituency service. Then, we show for the first time that these priorities map onto legislator behavior, such that the percentage of a district that is African American or Latino attenuates the responsiveness of legislators’ roll call voting to national variation in district conservatism and boosts the share of federal contract money that the district receives. In contrast, higher district per capita income amplifies the effect of district conservatism on
1
legislator roll call voting, but depresses attention given to constituency service. We also show that these effects are magnified in competitive districts, which speaks to the directionality of the relationship between priorities and representation. As we discuss in our conclusion, these findings show that elected officials are attentive to constituents’ priorities about spheres of representation and suggest one mechanism by which some groups come to be advantaged in the policy process (Bartels 2008; Gilens 2005; Griffin and Newman 2008).
Background Because “time is a House member’s scarcest and most precious resource,” MCs must carefully decide how they will allocate their energies while in office (Fenno 1978, 34). If MCs are “single-minded seekers of re-election,” they will devote greater time to activities that will maximize their re-election chances (Arnold 1990; Kingdon 1989; Mayhew 1974).1 Eulau and Karps (1977) identify four different types of “responsiveness” to constituents that MCs must prioritize: policy, allocation, service, and symbolic responsiveness. Policy responsiveness is reflected in the congruence of constituent policy preference and legislator action; allocation responsiveness is reflected in legislator success in distributive politics; service responsiveness is reflected in the legislator’s effectiveness aiding constituents in their personal interactions with government; and symbolic responsiveness is reflected in publicized gestures intended to
The specific mix of activities that will maximize an MC’s re-election chances is open to debate. For example, scholars have examined the impact of conducting casework on incumbent electoral support (Cain, Ferejohn, and Fiorina 1987; Fiorina 1989; Johannes 1984; Johannes and McAdams 1981; Krasno 1994; Serra and Moon 1984) as well as the impact of policy responsiveness (Ansolabehere, Snyder, and Stewart 2001; Canes-Wrone, Brady, and Cogan 2002; Erikson 1971; Johannes and McAdams 1981; Wright 1978). One reason that various studies have reached differing conclusions may be that citizens’ priorities vary across congressional districts, making no single activity uniformly most important for securing re-election. 1
2
strengthen constituency support and trust. Re-election minded MCs must discern how their constituents prioritize these spheres of representation. There is evidence that citizens as a whole do not value these spheres of representation equally (Kimball and Patterson 1997). Using the 1990 Senate Election Study, Krasno (1994) found that citizens hold varying priorities about what activities are most important for Senators to engage in. Among casework, working on national legislation, and securing funds for the district, citizens ranked working on national legislation as most important while doing casework for constituents ranked third. Cain, Ferejohn, and Fiorina (1987) found that constituents valued “keeping in touch with people about what the government is doing” most while “working in Congress on bills concerning national issues” ranked second and getting federal money for the district ranked third. In contrast, Grant and Rudolph (2004) found that working on local concerns was most important, national concerns less so, and helping people with personal problems with government was least important. Some have sought to map these priorities onto disaggregated and politically relevant demographic groups (Patterson, Hedlund, and Boynton 1975). For example, it appears that economically disadvantaged citizens and racial minorities tend to have few expectations of elected officials at all, and when they do these priorities tend not to focus on officials’ policy choices through roll call voting (Carman 2007; Davidson 1970). In addition, African Americans are more likely to expect their MCs to focus their energies on making sure their district receives its fair share of federal appropriations rather than working on national issues or conducting casework (Tate 2003, 128).2 Finally, a recent study showed that compared to whites, African Americans’ ideological distance from their MC is a weaker predictor of incumbent evaluation 2
Carman (2007) also found that African Americans are more likely to prefer a representative with a “trustee” orientation as compared to a “delegate” orientation.
3
and electoral support, which suggests that African Americans’ prioritization of policy responsiveness may differ from that of whites (Griffin and Flavin 2007). However, we still lack a comprehensive picture of groups’ prioritization of various aspects of representation (especially including policy responsiveness) and the consequences of this variation for congressional representation. It is likely that group differences in priorities about spheres of representation are rooted in the varied life experiences of groups. Economically and politically disadvantaged groups may place more weight on allocation responsiveness and constituent service because these groups’ relationship to government is more likely to involve distributive benefits (Jewell 1982, 145).3 Citizens who rely more heavily on government assistance and/or who are focused on making economic ends meet are likely to be more concerned about tangible benefits for themselves and their district and place less emphasis on whether their more abstract policy preferences are reflected in their MC’s voting behavior. Stated in political economy terms, Dixit and Londregan (1996) suggest that society’s poorest members are more likely to vote for “machine” candidates who principally focus on distributive activity because the poor value small private goods more than the wealthy do. In contrast, more politically informed groups, such as those with higher incomes or educational attainment, might be expected to place greater weight on policy responsiveness, since they have invested in obtaining policy-relevant knowledge, feel more competent to provide policy input to political leaders, and are better prepared to monitor elected officials’ implementation of their preferences (Arnold 2004; Delli Carpini and Keeter 1996; Hutchings 2001, 2003; Wolpert and Gimpel 1997).
3
For example, in his study of Aid to Families with Dependent Children (AFDC) recipients, Soss (1999) found that recipients’ general evaluations toward government bureaucracy are strongly influenced by their frequent interactions with AFDC service providers.
4
Are differences in priorities consequential to the democratic process? One study showed that citizens’ approval ratings of their Member of Congress are partly a function of their MC’s investments in the areas of casework and bill sponsorship, so long as the citizen also prioritizes these activities (Grant and Rudolph 2004). Stated another way, citizens seem to evaluate their MCs based on the perceived congruence of constituency priorities and legislator activity. This finding shows that there are strong electoral incentives for MCs to be responsive in their behavior to citizens’ heterogeneous priorities. As Grant and Rudolph (2004, 442) put it, “members must be mindful of what their constituents’ job expectations are and do their best to act in compliance with them.” To date, however, there is only fragmentary evidence that legislators are actually responsive to their constituents’ priorities about spheres of representation. In surveys, state legislators report that they are aware of constituents’ demands regarding casework and monetary allocations (Ellickson and Whistler 2001; Johannes 1984). Others have shown that state elected officials that represent constituents with a lower mean socioeconomic status report more requests and time spent on constituency service (Jewell 1982, 145-46; Thomas 1992). At the national level, racial minority MCs report that their constituents place special value on casework and respond accordingly (Thomas 1992).4 So, there is scattered evidence that elected officials are aware of the different types of responsiveness that their constituents value most. However, we are not aware of any study to date that explicitly relates citizens’ priorities about spheres of representation to their legislators’ actual behavior. We contribute to these literatures in a number of ways. First, whereas past studies have looked at variation in citizens’ priorities about spheres of representation or legislators’ 4
At the state level, Ellickson and Whistler (2001, 561-62) found that the race of a legislator does not predict time spent on casework but does predict time spent securing pork for constituents.
5
impressions of these priorities in isolation from one another, we examine the nexus of citizens’ priorities and MCs’ in-office behavior. This allows us to observe if legislators in fact allocate their efforts in a manner that reflects their constituents’ priorities. Second, we document variation in citizens’ priorities concerning roll call voting (i.e. policy responsiveness) which is important because (as we discuss below) the data used in prior studies has not asked citizens to explicitly choose between policy responsiveness and other spheres of representation. Third, we explore whether the effect of citizens’ priorities concerning spheres of representation on legislator action is conditional on district competitiveness. In doing so, we not only learn more about the ways in which priorities affect representation, we also bring evidence to bear on a thorny causal question that others have not confronted – whether citizens’ priorities about spheres of representation are rooted in the type of representation they have experienced.
Data and Method We seek to uncover (1) whether citizens place different priorities on various spheres of representation, and (2) whether these differences in priorities relate to differences in the way MCs allocate their time and effort. To measure differences in priorities about spheres of representation across demographic groups, we use a unique set of data from the 2006 Cooperative Congressional Election Study (CCES), an internet survey conducted in the preelection wave of the 2006 national election.5 We employed the standard lead-in used in the
5
The Common Content sample for the CCES is a nationally representative sample. Interviewed respondents were selected from the Polimetrix PollingPoint Panel using sample matching. A random subsample of 36,501 was drawn from the 2004 American Community Study (ACS), conducted by the U.S. Bureau of the Census, which is a probability sample of 1,194,354 with a response rate of 93.1% (participation in the ACS is mandatory). For each respondent in the selected ACS sub sample, the closest matching active PollingPoint panelist was selected using the following measure of distance: d(x,y). Following matching, the sample marginals were raked to the ACS marginals for age, race, gender, and
6
American National Election Studies (ANES) 1988-1992 Pooled Senate Election Study. Namely, CCES respondents were asked the following: “Here is a list of some activities that occupy U.S. House Representatives in Congress as part of their job. We want to know how important you think these activities are. Please rank these activities in order of importance by dragging them to the appropriate box,” where the destination “boxes” were labeled “most important,” “next most important,” and “least important.”6 The (randomly ordered) response alternatives were “Learning about district constituents’ opinions in order to better represent their views;” “Making sure the district gets its fair share of government money and projects;” and “Helping people who have personal problems with the government.”7 This item forced respondents to prioritize each form of representation as most, next, or least important, which some but not all related survey items have done (Grant and Rudolph 2004). The response alternatives for this item also differ from those in prior surveys such as the ANES Senate Election Study used by Krasno (1994) and Tate (2003) by explicitly offering policy responsiveness as one of the alternatives. We then ask whether citizens’ priorities systematically vary based on respondents’ race and income. After assessing whether various demographic groups tend to have different priorities about spheres of representation, we then test whether this has consequences for how MCs represent their constituents. First, we ask whether legislators are attentive to the level of policy
education. Raking was performed using iterative fitting. For more information on sample matching see Rivers (2006). 6
This item was not repeated for Senators, so our analysis focuses on representation in the House of Representatives. 7
The stem of the question alerts respondents to the trade-offs inherent to the activities in which Members of Congress decide to engage. The first response alternative was formulated by the authors. It is intended to measure citizens’ priorities about their MC’s attentiveness to their policy opinions. It is intended to highlight that representing citizens’ policy preferences requires a time commitment on the part of the Member. The second and third response alternatives are drawn directly from the ANES Senate Election Study.
7
responsiveness their constituents’ expect by seeing if district demographics moderate the level of policy responsiveness to district preferences. To perform these analyses, we require measures of district demographics, MC roll call behavior, and district preferences. To measure district demographics, we use 2000 Census data that has been updated to reflect the characteristics of congressional districts after the 2002 round of redistricting. To measure MCs’ general roll call behavior, we use their first dimension W-NOMINATE coordinates from the 109th Congress (2005-06) which range continuously from –1 to +1 with higher scores indicating a more conservative voting record (Lewis and Poole 2004).8 To measure district preferences, we follow prior studies in using the district’s 2004 two-party presidential vote share (Ansolabehere, Snyder, and Stewart 2001; Canes-Wrone, Brady, and Cogan 2002; Rothenberg and Sanders 2000). This measure equals the two-party percentage of the district popular vote received by George W. Bush in the 2004 election, normalized to reflect each district’s deviation from the national mean vote share for President Bush in 2004 (50.5%). We are also interested in the relationship between district demographics and another type of responsiveness: allocation. Following Stein and Bickers (1994), we focus on discretionary federal spending in congressional districts in order to avoid the complicating factor that some districts are needier than others of federal dollars. Our source for discretionary federal spending is the fiscal 2005 data on federal contracts as reported by OMB Watch and gathered from the Federal Procurement Data System. Following Stein and Bickers (1994), we model the percentage of all federal contract dollars received by a congressional district for the period October 2004 to October 2005. Specifically, we regress district contract percentages on district
These scores are usually interpreted as measuring a legislator’s social welfare voting record. Studies of policy representation have frequently employed this and alternative versions of NOMINATE scores as a dependent variable (e.g., Ansolabehere, Snyder, and Stewart 2001; Bartels 2008). 8
8
demographics, controlling for MC race and ethnicity, MC party affiliation, and MC seniority (Balla, Lawrence, Maltzman, and Sigelman 2002). Finally, we are interested in the relationship between district demographics and the degree to which MCs prioritize constituent services. To measure MCs’ efforts to focus on constituency service, we employ the number of district offices each MC maintains. We use the number of home offices because these sites, as opposed to an MC’s office in Washington, are primarily focused on fielding and responding to constituency concerns and requests. Members who maintain more offices (controlling for district geographic size and population) make it easier for constituents to access and interact with their district staff.9 We regress the number of home offices on district demographics, controlling for district geographic size and population.
Results To begin, we assess whether differences in respondents’ priorities about spheres of representation are related to race and income differences. Figures 1(A) and (B) report the percentage of respondents who prioritized each aspect of representation as “most important,” by race and income category. For every racial and income group we study, over half of the respondents ranked policy representation as the most important aspect of representation. Therefore, the y-axes in Figures 1(A) and (B) begin at 50% to more clearly highlight where the differences in representation priorities across demographic groups lie. Panel (A) reveals that African Americans and Latinos place less emphasis than whites on their MCs learning the opinions of constituents to better represent them but place more emphasis
Similarly, to measure Member attentiveness to “home,” Fenno (1978) employed the percentage of staff allocated to district operations. Fenno quotes a Member of Congress describing his allocation decision thus, “If you have a big staff, you create expectations, you encourage people to come to you and you get a huge case load” (1978, 49). 9
9
than whites on their MC making sure the district receives federal funds and on constituency service. Specifically, while 67.7% of whites place the highest priority on policy representation, only 61.9% and 60.5% of African Americans and Latinos do so, respectively. In contrast, while 29.6% and 26.2% of Latinos and African Americans prioritize allocations highest, just 18.1% of whites do so. Finally, while all groups place the lowest priority on constituent service, whites (14.1%) are more likely than African Americans (11.9%) and Latinos (9.9%) to do so. Part (B) shows that citizens with lower incomes prioritize policy responsiveness less than wealthy citizens and instead place more emphasis on constituent service. Specifically, while 74.6% citizens with household incomes of $70,000 or more place the highest priority on policy representation, just 60% of citizens with household incomes less than $40,000 do so. Instead, citizens in the lower income group (20.4%) are more than twice as likely as citizens in the high income group (8%) to prioritize constituent service. The prioritization of allocations differed little across the income groups.10 As an initial matter, citizens’ prioritizes about different aspects of political representation differ in important ways along racial and income lines. Next, we ask: Do elected officials respond to these different priorities about spheres of representation?
10
We also estimated a series of probit models to determine if these demographic differences are statistically significant and to account for various confounding factors. In these models, a binary variable indicated whether a respondent identified a sphere of representation as most important (1=Yes, 0=No). The results of these models showed that even after accounting for educational attainment and gender, African Americans (p=.10) and Latinos (p